> why the threshold that chunkErrorStrike to retire a block is setted
> as 3?
> ----
> Regards,
> Ezio.
>From my understanding this was developed back when large-geometry 1-bit
ECC SLC NAND was the norm. There a block wouldn't exhibit ECC errors
(correctable or not) until the block was quite close to worn out. In
that case it made sense to hanlde a couple of errant correctable ECC
errors but still retire the block after three.
But newer small-geometry SLC/MLC parts are more prevalent to returning
correctable ECC errors long before the block is near end of life and can
lead to premature retirement of blocks. You could try cranking up the
strike count from 3 to a higher value, but only if you undersand your
NAND's characterization to make sure its high enough that blocks don't
retire too early, but low enough to retire blocks that are at/near their
wear limit. Unfortunately some current NAND parts don't show any
correlation betyween corretable ECC errors and the block's wear limit to
allow properly setting the strike count limit...
--
Peter Barada
peter.barada@logicpd.com
(text/plain)